Show HN: NanoWakeWord – Open-source wake word training for any device

Category: ai-ml

Tags: wake-word, audio-processing, edge-ai, machine-learning, voice-interface

Score: 7.3/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 7)

NanoWakeWord is an open-source framework for training custom wake word models optimized for edge devices, featuring an automated ML engine that analyzes data and hardware to configure optimal architectures and training pipelines. It's interesting because it democratizes wake word creation by supporting multiple neural architectures (DNN, LSTM, Transformer, etc.) and includes phonetic adversarial negative generation to reduce false positives.

Target audience: embedded developers, ai engineers, iot developers, voice-interface developers

Repository: https://github.com/arcosoph/nanowakeword · Python · Apache-2.0 · 56 stars

View on Hacker News